OpenAI President Greg Brockman is urging enterprise security leaders to adopt AI-driven agentic systems to defend against evolving cyber threats, despite criticism regarding the company's own role in risks.

Key Points

• OpenAI President Greg Brockman advised CISOs to integrate AI agents into security workflows to identify and remediate system vulnerabilities before attackers exploit them.
• The recommendation follows the Hugging Face incident, which Brockman admitted demonstrated that OpenAI underestimated the real-world cyber capabilities of its own AI models.
• Suggested defensive tools include OpenAI’s Codex and the Codex Security plugin, which provide automated assistance for code review, vulnerability analysis, and supply-chain risk management.
• Industry analysts and security experts criticized the advice as self-serving, noting that OpenAI is actively selling the solutions to problems its own technology helped create.
• Critics, including CISOs from LexisNexis and Aikido Security, argued that OpenAI should prioritize accountability, liability, and "undo" mechanisms for rogue agents over sales-driven pitches.

Why it Matters

The push for AI-driven security reflects a broader industry shift where vendors are prioritizing cybersecurity capabilities to reassure investors ahead of potential public offerings. While these tools offer efficiency, they place the burden of risk management on enterprises, raising significant questions about vendor liability and the long-term safety of autonomous systems.